Re: Byte ranges -- formal spec proposal

Brian Behlendorf (
Fri, 19 May 1995 05:38:55 +0500

On Thu, 18 May 1995, Marc VanHeyningen wrote:
> (Aside -- apart from text/plain, how many widely used content-types
> are there where a fragment of the object is a legal object of that
> content-type?)

Absolutely no support for fragmentation: MPEG audio and video (I realized
this after using MPEG files as an example in my post last night - ack) I
would be willing to bet most other recent formats fit this description.
Also: GIF, JPEG, others. Most of these formats have smart readers which
can *recover* at some point into the stream (sorta like when you turn
your TV on and it's fuzzy for a second or two), so "no support for
fragmentation" doesn't mean "a subpart doesn't mean anything".

Minimal support for fragmentation (i.e., there are places where the object
can be split): PDF, Quicktime (certain codecs), HTML (?),
"mailbox-message" format plain text.

Completely fragmentable: plain text, aiff, wav, and mu-law sound formats.
Basically, any unstructured and uncompressed file formats, which also
happens to be the least useful type of file.

To repeat: just because a file's fragments aren't valid file types on
their own doesn't mean they're not useful. They might be valid in the
context of some other file type which has inlined them. For example,
let's say I have an HTML file like this:

An example of VRML:
<A REL="EMBED" HREF="http://host/foo.wrl?byterange=22342-22401"></A>

Whereas the object representation the URL pointed to was not a valid VRML
file, and thus was not typed VRML, but was perfectly embeddable in HTML.



--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-- http://www.[hyperreal,organic].com/